Top-down (descent) parsing gets a nice speed boost when you realize creating an AST isn't always necessary. Top-down parsing happens in the same traversal order

Top-down (descent) parsing gets a nice speed boost when you realize creating an AST isn't always necessary. Top-down parsing happens in the same traversal order as AST iteration, and this makes creating an AST not required. Bottoms-up parsing on the other hand typically requires creating the AST first, only to traverse it top-down in a manner resembling recursive descent. The speed advantages of bottoms-up are often lost in the inefficiencies of the AST. Of course this doesn't always apply. It depends on the language.